INDEX

Explanations

references to geographical locations, specifically the "District of Columbia."

oai_token-act-pair · gpt-3.5-turbo

occurrences of the phrase "District of" followed by a number or related context

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.11.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 predictive

-0.68

ocular

-0.66

urious

-0.66

behind

-0.64

eus

-0.64

Behind

-0.62

nery

-0.61

 misunder

-0.61

stellar

-0.60

 behind

-0.60

POSITIVE LOGITS

 Columbia

1.40

 Colombia

1.08

 Delaware

1.07

 Missouri

0.99

 Connecticut

0.97

 Georgia

0.96

 Illinois

0.96

 Idaho

0.96

 Maine

0.96

 Massachusetts

0.95

Activations Density 0.026%

references to geographical locations, specifically the "District of Columbia."

occurrences of the phrase "District of" followed by a number or related context

No Comments

No Known Activations